E2E Refined Dataset

E2E Refined Dataset是由奈良先端科学技术大学院大学创建的，旨在改善MR-to-text系统的质量。该数据集包含40,560条训练样本，用于自然语言生成（NLG）任务，特别是在餐厅推荐领域的MR-to-text应用。数据集通过手动注释和错误修正，移除了原始E2E数据集中的删除、插入和替换错误。此外，数据集还进行了文本和MR的细化，包括错误纠正、规范化处理和MR值的重新标注，以支持更精确的MR-to-text模型训练。

E2E Refined Dataset was developed by Nara Institute of Science and Technology to improve the quality of MR-to-text systems. This dataset contains 40,560 training samples for natural language generation (NLG) tasks, specifically MR-to-text applications in the restaurant recommendation domain. It eliminates deletion, insertion and substitution errors present in the original E2E dataset through manual annotation and error correction. Furthermore, both the text and MR components of the dataset have been refined, including error correction, normalization, and re-annotation of MR values, to support more precise MR-to-text model training.